In 2023, in Belgium, 101 people aged 50 to 69 spent several days in a lab where their brains were scanned, their sleep was recorded, and they were asked to solve memory and attention tasks. That is how the COFITAGE dataset was born, describing each participant from five angles: molecular, structural, physiological, genetic and cognitive. These layers used to be studied separately, but here they are combined into a single portrait.
Why does this matter? Alzheimer's disease begins decades before the first memory lapses. By age 50-69, healthy people already show subtle changes, but catching them one by one is nearly impossible. Large databases such as ADNI or OASIS are disease-oriented and provide coarse phenotyping, while deep studies are usually small and fragmented. COFITAGE stands apart: it is the first open cohort to combine PET, advanced MRI, polysomnography, APOE genotype and detailed neuropsychology in the same people within a key age window.
How were the data collected? Participants underwent amyloid PET (100 people) and tau PET (67 people) to see the buildup of proteins linked to Alzheimer's. MRI at 3 Tesla provided high-resolution anatomical images, multiparametric maps and diffusion with three shells, which reveals the course of nerve fibers. Two nights of polysomnography with 11 EEG leads recorded sleep architecture, and genotyping determined APOE ε2/ε3/ε4 alleles. A neuropsychological battery of 15+ tests covered attention, executive function, memory, language and cognitive reserve. All data were converted to BIDS and checked with a validator.
What is already visible? The dataset includes 101 participants with a full set of modalities at baseline and 65 people with repeat cognitive testing after two years. Amyloid PET was performed in 100 people, tau PET in 67, and both modalities are available in 66. MRI includes T1-weighted images, T2-weighted hippocampal slice images with a voxel of 0.4×0.4×1.2 mm³, multiparametric maps and diffusion. Polysomnography was recorded over two nights for each participant, with automatic scoring of sleep stages in 30-second epochs. APOE genotype was determined in everyone, and the allele distribution matches the population.
What is the value? COFITAGE differs from neighboring open cohorts not in scale but in depth: a single sample combines molecular, structural, physiological and cognitive layers that are usually studied separately. This makes it possible to test hypotheses about the interplay of amyloid, tau, sleep and cognitive reserve at the individual level, not just on group averages. The work does not unveil a new biomarker, but it lays the foundation for such discoveries: these data have already shown that cortical reactivity and daytime fluctuations in brain activity are more sensitive to early aging than static measures.
What comes next? The dataset will keep growing: upcoming releases will include raw polysomnography recordings, EEG, EEG-TMS, saliva and urine samples, actigraphy, and a seven-year cognitive follow-up for roughly 65 participants. This will allow a shift from cross-sectional associations to longitudinal trajectories and make it possible to test which baseline measures predict individual changes after seven years. The open BIDS structure invites the development of new methods for multimodal fusion and machine learning on deeply phenotyped but small samples. And so the portrait of the aging brain gains depth, letting us discern the first strokes of changes to come.
COFITAGE participants slept in the laboratory two nights in a row to rule out the first-night effect, and their sleep during the week before recording was tracked with an actigraph and a diary. Such a rigorous protocol is rarely found in neuroimaging cohorts.
Not published on arXiv. Here are its page, the PDF and the formalization: you can check it yourself, not only read our account.